Genomic methylation patterns are required for normal X chromosome inactivation, genomic imprinting, and the repression of parasitic sequence elements. Large reductions in methylation levels are lethal to mammalian embryos. Methylation patterns are created by a complex series of demethylation and de novo methylation events; demethylation occurs in primordial germ cells and preimplantation embryos, while waves of de novo methylation occur during gametogenesis and early postimplantation development. Tight regulation of de novo methylation is essential, as errors in the process can inactivate essential genes such as tumor suppressors or lead to biallelic expression of imprinted loci. However, almost nothing is known of the mechanisms that regulate de novo methylation. We have identified a new family of DNA methyltransferases that are strong candidates for the agents that establish methylation patterns. These enzymes share catalytic motifs with all other DNA (cyto-sine-5)-methyltransferases but are most similar to a family of very unusual phage-encoded enzymes that each recognize several different sequences via modular recognition domains that swing in and out of a conserved catalytic core structure. The new mammalian enzymes, which we have named Dnmt3alpha and Dnmt4alpha, are found in embryos and are the strongest candidates yet for the long-sought de novo methyltransferases. DNMT3alpha is likely to be involved in a lethal human disease. ICF syndrome is the only genetic disorder that involves global abnormalities in methylation patterns, and DNMT3alpha has been mapped to the small region of human Chr 20p that contains the ICF disease locus. It is very likely that DNMT3alpha is the ICF genes, and we will test this hypothesis by determining whether homozygous mutations in DNMT3alpha are present in ICF patients. Dnmt3alpha and Dnmt4alpha will be expressed in heterologous cell types and tested for in vitro enzymatic activity and sequence specificity through a very sensitive and specific assay that involves formation of covalent complexes with a mechanism-based inhibitor. The sequence and organization of Dnmt3alpha and Dnmt4alpha are very different from known bacterial and mammalian DNA methyltransferases, and it will be very interesting to learn if they are the first metazoan DNA methyltransferase whose sequence specificity extends beyond the CpG dinucleotide. Analysis of the new family of mammalian DNA methyltransferases is almost certain to give important insights into the epigenetic mechanisms that mammals use to restrict the future transcriptional potential of specific promoters.